Airsys Pushed for New Cooling Metrics in AI Legislation
The firm seeks to standardize efficiency reporting as data center rack power demands climb toward 600kW.
Updated on Sept. 22, 2026 in Data Centers

Live Poll
Should the government mandate energy efficiency standards for the expansion of data centers?
Infrastructure firm Airsys has advocated for the inclusion of a Power Compute Effectiveness metric within the proposed Liquid Cooling for AI Act. The push comes as surging rack densities threaten to overwhelm existing grid capacities in regions like Texas.
Why it matters
Rising rack densities are creating a performance discontinuity that limits data center scaling. Standardization is intended to address grid capacity constraints by curbing the 22MW of power currently wasted by legacy cooling systems per every 100MW site.
Airsys reports that switching to a two-loop cooling architecture improves Power Compute Effectiveness from 0.73 to 0.95. This upgrade allows for a 30 percent increase in compute capacity from the same grid connection.
The players
Airsys
An infrastructure company specializing in data center cooling technologies and thermal management systems.
NVIDIA
A semiconductor company that designs high-performance graphics processing units and data center architectures.
Uptime Institute
A research and advisory organization that benchmarks data center performance and uptime standards.
AFCOM
An association for data center and mission-critical infrastructure professionals that publishes industry data.
The details
Data centers increasingly rely on high-density hardware, with NVIDIA Rubin Ultra systems reaching 600kW per rack, far exceeding the current 27kW average reported by AFCOM. Airsys proposes replacing legacy cooling with two-loop architectures, which use a primary fluid loop to transfer heat to the exterior and a secondary, internal loop to dissipate heat directly from high-performance components. This approach reduces the reliance on power-intensive air-handling equipment that currently accounts for substantial energy losses.
Timeline
November 2025: The Liquid Cooling for AI Act was introduced.
April 2026: The Energy Subcommittee held a hearing on the act.
September 2026: Airsys officials met with congressional offices.
The Tech Race
The push for new cooling metrics follows the introduction of the Liquid Cooling for AI Act in November 2025. It marks a critical attempt to align infrastructure cooling efficiency with the rapid power scaling of next-generation hardware like the Rubin Ultra.
The proposed regulations could force data center operators to adopt more efficient cooling architectures to comply with grid-load requirements. This shift is essential to prevent electricity demand from increasing by 125 percent in high-growth areas like Texas.
The takeaway
The industry is moving toward rack densities that legacy cooling systems can no longer support efficiently. Stakeholders should monitor the Energy Subcommittee for the next legislative session updates regarding the Liquid Cooling for AI Act.
Further reading
Learn more about the latest innovations in Data Centers infrastructure.
Source note: This article includes information reported by Capacity.
Live Poll
Should the government mandate energy efficiency standards for the expansion of data centers?









